Neurodegenerative Disease Detection using Deep Convolutional GANs and CNN

Abstract

Over the past ten years, advances in deep machine learning techniques, high-speed computing infrastructure development, and an improved understanding of deep learning algorithms have created new opportunities for advanced analysis of neuroimaging data. Neuroscientists can now use the data from neuroimaging to diagnose neurodegenerative diseases. Yet, due to the similarities in disease characteristics, it is challenging to identify such disorders from neuroimaging data accurately. The reason for such results is the current or inevitable limited availability of neuroimaging data. Thus, this paper suggests an unsupervised generative modeling technique using Deep Convolutional Adversarial Networks to produce synthetic images (DCGANs). This method uses the ADNI dataset, which contains data for four neurodegenerative diseases, namely: Alzheimer’s Disease(AD), Mild Cognitive Impairment(MCI), Early Mild Cognitive Impairment(EMCI), Late Mild Cognitive Impairment(LMCI), and subsequently uses DCGAN on the small quantity of data, so increasing the dataset’s size and variety by utilizing GAN. To outperform the conventional deep learning techniques, the artificial images, and the original dataset images are combined and trained into a convolutional neural network (CNN).